Ali Ghadirzadeh

About

Ali Ghadirzadeh is a robotics and machine learning researcher whose work centers on deep reinforcement learning, robot skill acquisition, and adaptive human-robot interaction. His most influential contribution, "Deep Predictive Policy Training using Reinforcement Learning" (2017, 113 citations), introduced a novel framework for training robots to learn skilled tasks by accounting for the inherent latency of sensorimotor processes — a foundational advance in robot learning. Building on this, Ghadirzadeh has tackled critical challenges in real-world robot deployment, including navigation in complex environments, sim-to-real transfer via meta-learning, and few-shot policy adaptation across different robotic platforms — addressing the costly problem of retraining from scratch whenever hardware changes. His work on adversarial feature training and affordance learning further advances generalizable visuomotor control, reducing reliance on large task-specific datasets. Notably, his research extends beyond purely technical domains into co-adaptive human-robot cooperation, even examining the neural correlates of human-robot coordination. Supported by Sweden's Foundation for Strategic Research through the COIN project, Ghadirzadeh's cumulative contributions reflect a researcher committed to making robot learning more efficient, transferable, and meaningfully integrated into human environments.

Research Focus

Key Achievements

Top Papers

1. 1
   Deep predictive policy training using reinforcement learning

   113 citations · 2017
2. 2
   Deep Reinforcement Learning to Acquire Navigation Skills for Wheel-Legged Robots in Complex Environments

   47 citations · 2018
3. 3
   Bayesian Meta-Learning for Few-Shot Policy Adaptation Across Robotic Platforms

   25 citations · 2021
4. 4
   Meta Reinforcement Learning for Sim-to-real Domain Adaptation

   17 citations · 2020
5. 5
   Co-adaptive Human–Robot Cooperation: Summary and Challenges

   16 citations · 2021
6. 6
   Deep Predictive Policy Training using Reinforcement Learning

   16 citations · 2017
7. 7
   Adversarial Feature Training for Generalizable Robotic Visuomotor Control

   14 citations · 2020
8. 8
   Back to the Manifold: Recovering from Out-of-Distribution States

   12 citations · 2022
9. 9
   Affordance Learning for End-to-End Visuomotor Robot Control

   11 citations · 2019
10. 10
    Coordinating With a Robot Partner Affects Neural Processing Related to Action Monitoring

    9 citations · 2021

Key Collaborators

Contact & Links